How To Automate Perplexity AI Web Search with Zapier
Let's build with Zapier and AI (100+ videos)📅 2025-10-07
Transcript
[00:00] Perplexity AI released a big update. We can now leverage its web search capability automatically in software. So, in today's video, we're going to see how we can use it in over 8,000 apps in Zapier's ecosystem. Pretty nice. Let's jump in. Today's video is sponsored by Zapier because as you already know, me and Zep have teamed up and they said, "Corbin, we got to do a bunch of cool stuff and Perplexity API just dropped its web searching capability. It's pretty intuitive." So, my intention is we're going to create a workflow. In this particular tutorial, you're going to learn the important skill of now knowing how to integrate internet into a workflow, accessing the internet, doing
[00:31] a search query, receiving data, and leveraging that. This is pretty cool stuff here. Let's do it. To start off here, we're going to go ahead and say create new zap in Zapier's AI orchestration here. All the different platforms, interfaces, tables, everything we love. Our trigger here, I'm going to set to auler. The main purpose of what I want to show you today is how to just integrate a web search capability with an automation which is pretty new, really cool, and we get that extra layer of AI in the search results. Now, one caveat I want to bring up is that this is very different than the traditional web parser that we had to
[01:01] use in the past due to the fact of the limitations that was associated with the web parser when parsing a web page. A really big limitation of using the internet and accessing the internet in this method was we had to give a specific web page. No fun. But using perplexity API, we can essentially have software just do my Google search for me or my Bing or maybe you're on the duck duck of the go. So we're going to schedule this event. What I want to do is I'm going to create an event that's going to follow a list of stocks that I care about. I'm going to see if there's any new news in regards to it. Give me a
[01:31] summary analysis and then have that either emailed, text, or slacked to me. So, we say every day because every day could be a new day for a stock. Time of day, we are going to go with 6 a.m. Choose 6 a.m. Continue. And then we'll test this trigger. First step done. We have our trigger. Let's go to Perplexity API. To get the API key you're going to need, you're going to simply come over to your account and go to API. Once you do that, you're going to come over here to API keys and we're going to say accept terms and generate API key. We're going to give it a name. So, I'm going to say my ZAP your key create key. Once I do that, you're going to see the secret key show up here. And in order to
[02:02] use it, we're going to copy here. So, that's the first step. Get your key. Let's get access to the capabilities of AI and internet searching. I like it. Next step, I'm going to leave this in the description down below. This is the docs relevant to sonar web search control, how it works with artificial intelligence. You can get more information this way. For now though, we're going to come back to our zap here. I'm going to rename it. I'm going to call it web search finally. And then we're going to add our block here. So, we're going to do perplexity AI. So, we add block as step. We're going to type in perplexity here. Boom. With perplexity open here, we can choose our event. We're going to say chat completion. And then we'll sign into our
[02:33] account. This is where we're going to get prompted to put in the API key here. And then we're going to say yes, continue perplexity. That's the key we copied earlier. Paste it there. Once we do that, we can go to configure here and begin. So, what's really nice about Perplexity here is that we have a couple different models to choose from. In most use cases, since we're just caring about web search, we're going to opt for sonar. If you're wondering what these other ones are, such as deep research, if you're familiar with other models like Chad GBT or Claude, that's when the model takes extra time. Eg. A way to equate that is that your automation would take longer because it's doing
[03:03] more search queries on the relevant topic you want. So for example, maybe if I use just sonar, that would take me 10 seconds for an output. But if I use deep reasoning, it could take upwards of 30 plus seconds. So keep that in mind. So for the user message here, in order to prompt the web searching capability, let's just straight up just say it. So here's how you want to prompt it. We're going to say, I want you to look at these stock tickers today. Web search Apple, Nvidia, Netflix, and those are the relevant tickers. Notice how in the actual prompt I'm identifying web search so the AI knows hey this isn't just a
[03:34] genuine query like what color is the sky rather we need up-to-date information therefore search the internet nice so for the system message this is be precise and concise as the default but what we're going to do here is we're going to say give me a four sentence summary of the news if there is any we can go crazy here as you know of AI and prompting we can really just do anything we would want and we could say is the news bearish or bullish English. Should I buy the stock right now? What price is the stock? This really just comes down to a very simple lesson that I like
[04:06] teaching anyone that wants to learn how to prompt AI. And it's this. If you want to understand the best way to do prompts in this context, think of the prompt like you're about to teach a human how to do that task. Eg. You have a new employee for your company. Here's your onboarding. Here is how you do the task. Step one, step two, step three, step four. This is good prompt structuring. Treat it like a human. Teach it like a human. And it'll give you pretty good outputs. For now, I'm going to keep it simple. It's just going to look up these three tickers, give me a four sentence summary of each one if there is any news. Now, we're hit continue and let's watch the magic test step. This is groundbreaking, y'all. We weren't able to search the internet like this this
[04:37] intuitively 2 years ago, a year ago. I love Perplexity API's search API. So, first thing you'll notice is it's going to give citations. We love that. We are used to the past when we don't even know if the information is true or not cuz it's just like saying stuff, but there's no citations. But now, we get that with all these relevant links. So, we can include that in my email here. Scrolling down, we can see all the different news articles that they found and the specific results URL. Apple historical stock price data, US market, live quotes, news ideas, trading view. So based off all these citations, we can scroll down here and what we're looking for is the assistant response.
[05:08] Everything prequel to the assistant response is just the data that's getting plugged and played in order to give you this output. So as we identified in our prompt here, we're looking for sentences if there's a news story of what we should care about. Right now it is October 1st, so this is nice. And as you notice here, it's actually working pretty effectively here in the sense that there was no news on Nvidia on October 1st that's relevant enough for it to show up and there was no update. Now, obviously the output here is a very general, but we could have made our prompt more specific as I identified
[05:39] before when it comes to bearish bullish, what you really care about or why you're actually searching the internet for your automation use case. Let's go ahead and send this to myself, though. We're going to add an action here and anything it integrates with, right? So, in theory, I could do SMS. I could text myself. We could send ourselves a nice little Slack message, but for me, we'll just do Gmail. Hit Gmail. Make sure your account's selected. We'll do the action event of send email. The two we'll just send to ourselves. Yeah, I'm talking to myself, so we'll send it to ourselves. So, for me, it's just contact at webbcafi.com. Scrolling down here, we can do a subject line of, you know, daily summary of stocks. Get as fancy as
[06:12] you want. Obviously, Zapier is great in the sense that we can basically do everything we would want to do with an email just through automation, right? Okay, we can add CC, BCC, the from name, body type, plain HTML. Coming down here to the body though, we're going to hit this little add icon so we can get data from the previous step. The data we're looking for here is check completion and specifically assistant response. Now, one thing I want you to know about Zapier, which is very helpful compared to other platforms I've seen is that it will prioritize the type of data that you see in previous blocks. So, it isn't just a bunch of jumbled information like this. It's like, wait, which one is it,
[06:42] Corbin? Which one is it? It's typically the one that's going to be at the top of the flow. So the one that we care about is going to be the assistant response which has all that information we requested in regards to those stocks signature. Continue and we're going to test this step. So we'll test that step and then coming over to the email. Boom. I went ahead and started already daily summary of stocks and then we have our systems output. This data though can be leveraged in any context that Zapier integrates with when it comes to applications on their ecosystem. Therefore, let me show you something really cool that you can start doing to come up with ideas that could possibly help you. And that really cool thought
[07:14] experiment you can do is that when using perplexity AI or Chad GBT web search or claude web search, is there reoccurring queries that you're putting into these chat bots that require the internet for the data to be coming out? Eg, I search in here, what's the weather today? In theory, we could set that up as an automation to find out what the weather is today. or as we just saw with the stock play. But the idea is that if you already actively use Perplexity AI, you can look at your past web search when it came to what you search actively on the internet and we can start automating that. That just about does it though. So make sure to leave a like. It's
[07:45] completely free. Now you know how to connect the internet [music] into your automations. No more Google searches. I'm done Google searching stuff. We can now let artificial intelligence do it for us with the help of Zapier automating everything [music] in the back end. And the best part is that you can start using Zapier for free. You can also get five free credits on Perplexity API as well. Those are two random videos. That is my face. And as you already know, I'll see you in the next [music] Zapier automation with internet